Nodelace

Nodelace renders polished architecture, flow, and sequence diagrams from a small text language. Rendering is deterministic and entirely local: the base package has no runtime dependencies, AI calls, network requests, telemetry, Graphviz, browser, or Node.js requirement.

python -m pip install nodelace

Create system.diagram:

architecture "Small Web Service"
direction left-to-right

Browser -> API: HTTPS
API -> Database: query
highlight API

Then validate and render it:

nodelace check system.diagram
nodelace render system.diagram

The default artifact is system.svg. Standalone HTML is also available:

nodelace render system.diagram --format html -o system.html

The Python API accepts source text or an immutable parsed model:

from nodelace import parse_diagram, render

diagram = parse_diagram('flow "Two Steps"\nStart -> Finish')
svg = render(diagram)

Nodelace provides:

• architecture, flow, and basic sequence diagrams;
• deterministic layered and participant-lane layout;
• orthogonal connectors, cycles, flat groups, and focal highlights;
• accessible SVG title and description metadata;
• self-contained offline SVG/HTML with bundled OFL-1.1 fonts and notices;
• a strict, non-executable UTF-8 language with bounded input limits;
• a dependency-free Python 3.11+ runtime and typed public API.

Current limits include flat groups only, Latin coverage in the bundled fonts, no full UML sequence fragments, a 50-node safety ceiling, and SVG/HTML output only. Other Unicode text remains valid and selectable but may use a viewer's installed fallback font.

The source distribution contains the complete language and architecture references under docs/, plus ten source examples, byte-verified SVGs, and an offline gallery under examples/.

Nodelace code is MIT licensed. Instrument Serif, Geist, and Geist Mono remain under the SIL Open Font License 1.1; their complete notices ship in the package and in artifacts that embed the fonts.

The design goals were inspired by Kathryn Lavery's diagram-design work. Nodelace is an independent deterministic implementation and does not package or invoke that AI-oriented skill.